Dahua’s AI achieved two breakthrough key indicators to achieve industry leadership
Recently, Dahua Co., Ltd. developed a weakly supervised semantic segmentation technology for pathological slices in the medical field based on a self-developed training framework, and won the first place in the WSSS4LUAD competition; the developed instance segmentation technology won the first place in the MSCOCO competition; key indicators surpass first-class AI The company and top academic research institutions demonstrate Dahua’s deep technical strength and innovation capabilities in the field of target segmentation.
The WSSS4LUAD (Weakly Supervised Semantic Segmentation for Lung Adenocarcinoma) competition and the LUAD_HistoSeg data set were initiated by the Guangdong Provincial People’s Hospital, University of Electronic Science and Technology of China, Beijing University of Posts and Telecommunications and other institutions and scholars. It aims to use image-level classification tags to achieve histopathological images Semantic segmentation is of great significance to promote the rapid development of artificial intelligence in pathological images.
The MSCOCO (Microsoft Common Objects in Context) data set is a large data set constructed by Microsoft. There are 80 categories of instance segmentation tracks. It is one of the authoritative data sets for measuring model performance. It has attracted well-known AI laboratories at home and abroad. Active participation in top academic research structures.
According to statistics, Dahua has obtained more than 60 firsts and more than 1,300 AI invention patents in the global AI competition evaluation. Dahua AI has achieved rapid breakthroughs and has always maintained its technological leadership. It is inseparable from the scientific research capabilities of industrial innovation endowed by Dahua Juling platform. The self-developed training framework of the Dahua Juling platform can efficiently and accurately complete Dahua’s more than 1 million training tasks each year, and the fastest completed task only takes 10 minutes, which effectively supports Dahua’s technical research in the field of new AI algorithms. .
The field of medical pathology realizes the recognition of tissue slices through semantic segmentation technology
This award-winning AI algorithm relies on the self-developed Dahua Juling artificial intelligence development platform, and uses weakly supervised and semi-supervised learning technology to greatly reduce the data requirements of the algorithm. It is used in medical pathology, security X-ray, millimeter wave imaging, etc. The domain solves the difficult problems of obtaining training materials and extremely high professional requirements, which greatly improves the accuracy of target segmentation and effectively helps the landing of scene-based applications.
Analytical effect of instance segmentation in traffic scene
Relying on instance segmentation technology, in the field of intelligent transportation, it can help automatically analyze traffic elements such as traffic signs, markings, signal lights, and man-machine non-humans. It is widely used in multi-scenario services such as electronic police, road bayonet, traffic incident detection, and smart parking. Realize effective identification of road traffic situation, traffic incidents, etc., and comprehensively improve urban traffic governance capabilities.
Nowadays, artificial intelligence applications are blooming, and the demand for intelligence continues to grow. Dahua Co., Ltd. will be guided by the strategy of “Dahua Think# Cloud Connecting Everything, Digital Intelligence, Future”, focusing on continuous innovation in urban and corporate business scenarios, from traffic management to people’s livelihood services, from safe production to intelligent manufacturing, from agriculture, animal husbandry and breeding to environmental protection. Constantly breaking through the boundaries of traditional business, continuously expanding AI application fields, and working with customers to build a leading edge in digital intelligence applications, empowering thousands of industries and hundreds of industries to upgrade digital intelligence, and promoting sustainable, healthy and high-quality economic and social development.